After model estimation, MPlus produces a .gh5 file that contains the data necessary for graphing (so they claim). The data extracted from a .gh5 file contains a column for the unique intercepts and slopes (so it seems). The trajectories produced with these values
yHat = B_Y + LINEAR*TIME
On the other hand, we can look at the fixed effect estimate in the .out file. Using the values for the estimated fixed effect we can produce the individual trajectories, conditional on the values of predictors
yCond = int + slope*TIME
So it is unclear which predicted trajectories are the correct ones. If .gh5 data can be trusted as the unique intercepts and slopes then there is a mistake in the computation of predicted trajectories from estimates in the .out file.
yHat : predicted value reconstructed from .gh5 estimates (int, slope) Semi-transparent blue lines: individual predictions Red line: Main effects
yCond: predicted values reconstructed from .out estimates (fixed effects) Semi-transparent blue lines: individual predictions Red line: Main effects